grinblattk_jf56

7/29/2019 GrinBlattK_JF56

1/21

How Distance, Language, and Culture

Influence Stockholdings and Trades

MARK GRINBLATT and MATTI KELOHARJU*

ABSTRACT

This paper documents that investors are more likely to hold, buy, and sell the

stocks of Finnish firms that are located close to the investor, that communicate in

the investors native tongue, and that have chief executives of the same cultural

background. The influence of distance, language, and culture is less prominent

among the most investment-savvy institutions than among both households and

less savvy institutions. Regression analysis indicates that the marginal effect of

distance is less for firms that are more nationally known, for distances that exceed

100 kilometers, and for investors with more diversified portfolios.

IT HAS LONG BEEN KNOWN that most investors shun foreign stocks in their

portfolios ~see, e.g., French and Poterba ~1991!, Cooper and Kaplanis ~1994!,

and Tesar and Werner ~1995!!. This phenomenon, known as home bias,

refutes the implications about investor behavior developed in many stan-

dard asset-pricing models. We do not know the root cause of home bias, nor

do we know if there are differences in home bias behavior across investors.

Researchers such as Stulz ~1981a, 1981b! and Serrat ~1997! have hypoth-esized that home bias may be due to restrictions on international capital

flows or the nontradability of some goods across international boundaries.

However, recent research suggests that home bias may be part of a larger

phenomenon in which investors exhibit a preference for familiar companies.1

* Mark Grinblatt is from the Anderson School at UCLA, and Matti Keloharju is from the

Helsinki School of Economics, Finland. We are grateful to the Finnish Cultural Foundation, the

Foundation of Economic Education, the UCLA Academic Senate, and the Yrj Jahnsson Foun-

dation for financial support and to the editor, Ren Stulz, two anonymous referees, Shlomo

Bernartzi, Kenneth French, Chris Gadarowski, Michael Long, Tim Loughran, Rena Repetti,

Emilio Venezian, Tuomo Vuolteenaho, and seminar participants at Chicago, MIT, Northwest-ern, Notre Dame, Rutgers, and Yale for comments on an earlier draft. We wish to thank Antti

Lehtinen and Jorma Pietala for the distance data, and Kari Toiviainen for suggesting the use

of annual reports as indicators of language. We are especially indebted to Henri Bergstrm,

Mirja Lamminp, Tapio Tolvanen, and Lauri Tommila of the Finnish Central Securities Deposi-

tary for providing us with access to the investor data. Some of this research was undertaken

while Grinblatt was at Yales International Center for Finance, for whose support we are grateful.1 Huberman ~1998! observes that Regional Bell Operating Companies are more likely to be

held by investors who subscribe to their local telephone service. Coval and Moskowitz ~1999a,

1999b! document that mutual fund managers prefer to hold locally headquartered firms and

hint that this may be due to easier access to information about the firm. Petersen and Rajan

~2000! study the effect of distance on lending relationships. Kang and Stulz ~1997! show that

Japanese firms with a greater international presence, as evidenced by having ADRs, or a

great deal of export business, have greater foreign ownership. Tesar and Werner ~1995! show

that U.S. investors exhibit a bias towards Canadian stocks in their foreign investment.

THE JOURNAL OF FINANCE VOL. LVI, NO. 3 JUNE 2001

1053


2/21

Familiarity has many facets. The firms language, culture, and distance

from the investor are three important familiarity attributes that might ex-

plain an investors preference for certain firms. This paper finds that all

three of these attributes contribute to investor preferences for certain stocks.It also shows that the preferences tied to these attributes are inversely re-

lated to investor sophistication.

The results in the paper, developed by analyzing the holdings, purchases,

and sales of Finnish stocks while controlling for numerous alternative ex-

planations, show that:

Investors in various municipalities in Finland prefer to hold and trade

stocks headquartered in nearby locations to those in more distant

locations.

The distance effect is piecewise linear in the log of distance with anabrupt change in the slope of the distance coefficient at 100 kilometers.

Firms that are headquartered in Helsinki, and thus are more nationally

known, have less of a distance effect associated with them.

Investors whose native tongue is Finnish prefer to hold and trade Fin-

nish companies that publish their annual reports in Finnish to Finnish

companies that publish their reports in Swedish and vice versa. Multi-

lingual companies lie between the one-language companies in the pref-

erences of both the Swedish-speaking and Finnish-speaking investors.

Controlling for the language the firm communicates in and the distance

from the investor, investors in Finland prefer to hold and trade firms

whose CEO is of similar cultural origin.

The inf luence of distance, language, and culture on stockholdings and

trades is generally smaller for financially savvy institutions than for

households or less savvy institutions.

The inf luence of distance and culture on stockholdings and trades is

smaller ~but still sizable! for more sophisticated household investors.

In contrast to most previous studies of home bias and related stock pref-

erences, our analysis focuses on open market purchases and sales, as well as

shareownership. There are several important reasons for this. First, dis-

counts are typically given to Finnish employees for IPO participation. Be-cause there is some possibility of a modest bias in the shareownership results

as a consequence of these IPO discountsemployeeparticipants in these

IPOs tend to live near the firms at which they workwe analyze open mar-

ket buys and sells in addition to shareownership. The buys, in particular,

exclude IPOs and gifts as a source of acquisition.2 In addition, there is a

2 A bias may exist in the sell ratios if investors rebalance their portf olios after partic ipating

in such IPOs. Employee stock ownership plans and stock options affect relatively few investors

in Finland and thus are unlikely to more than negligibly bias the results on shareownership.

Moreover, we have reanalyzed our data excluding all investors who live in the same munici-

pality as the company headquarters. Although this eliminates a large fraction of the employees

of the company as potential shareowners or traders, our results are largely unchanged.

1054 The Journal of Finance


3/21

potential feedback effect in the shareownership results. If a company per-

ceives that a large proportion of its shareowners prefer a particular lan-

guage, the company may choose to communicate in that language. This

feedback effect is not present when analyzing buyers of the companys stock.Finally, we are interested in buys and sells in addition to shareownership

simply because it is interesting to see whether investors who live near a firms

headquarters municipality are more actively buying and selling that stock.

The paper is organized as follows. Section I of the paper describes the

data. Section II presents the results. Section III concludes the paper and

provides thoughts on the implications of the results for corporate policy and

future research.

I. A Unique Dataset

A. Motivation for and Description of the Dataset

Restricting our focus to intracountry investment behavior simplifies theanalysis of investor preferences in that restrictions on capital flows and the

intercountry nontradability of some goods cannot explain or confound our

findings.3 We chose Finland for several reasons. First, the Finnish Central

Securities Depositary ~FCSD! maintains daily comprehensive official records

of shareownership and trades in electronic form and has provided us with

access to approximately two years of historical data f rom these records.4 The

data analyzed include the FCSDs January 1, 1997, shareownership records

and all trades between December 27, 1994, and January 10, 1997, for its 97publicly traded companies.5

Language differences also make Finland interesting to analyze. There are

two official languages in Finland: Finnish and Swedish. Finnish speakers

account for 93 percent of the population, whereas Swedish speakers account

for 6 percent of the population. However, the inf luence of the Swedish-

speaking investors in the Finnish financial markets exceeds what their frac-

tion of the population would suggest. At the beginning of 1997, for example,

Swedish speakers held 23 percent of household shareowner wealth.

Finnish companies also exhibit language differences. Some Finnish firms

communicate exclusively in Finnish, others communicate exclusively in Swed-ish, and still others communicate in multiple languages, typically Swedish

3 Although some goods may be less tradable within Finland than others, the shorter dis-

tances, unified financial markets, homogeneous regulatory environment, and lack of restric-

tions on labor, consumer, and capital mobility make this issue a rather negligible one in comparison

with the international nontradability of goods.4 See Grinblatt and Keloharju ~2000, 2001! for details on this dataset. The register is the

official ~and thus reliable! recording on a daily basis of the shareholdings and trades of virtu-

ally all Finnish investorsboth retail and institutional. In contrast to Finnish domestic invest-

ment, the data on foreign investment in Finnish stocks that we employ is not comprehensive.5 Sixty-two of these are headquartered in Helsinki. One company, with fewer than 100 share-

holders, was excluded because of its exceptionally small size. Five companies were delisted over

the two-year sample period, leaving 92 firms for the January 1, 1997, shareownership analysis.

Distance, Language, and Culture 1055


4/21

and Finnish, Swedish, Finnish, and English, or Finnish and English. Be-

cause the larger companies tend to be multilingual, our analysis of language

~and distance! effects use firm dummies to control for confounding firm

attributes, like firm size. Also, the language of the company may differ fromthe cultural background of senior management, and the cultural background

of senior management differs across multilingual firms, allowing us to dis-

tinguish language from cultural preference.

Prssitieto 1995 and 1996 report the municipality in which Finnish firmsare headquartered, the name of the CEO, as well as the language of the

companys annual report.6 We classify the language of Finnish firms as the

language of their annual reportSwedish, Finnish, or multilingualwhich

is generally the language of the other financial information reported by the

firm. Of the 97 firms, 2 report only in Swedish, 12 report only in Finnish,

and the other 83 are multilingual.7

We classify the culture of the firm basedon the name and native language of the CEO.8 Eighty-three of the firms are

of Finnish cultural origin; the remaining 14 are of Swedish cultural origin.

No firms have CEOs that are not of Finnish or Swedish cultural origin.

The FCSD database either contains or can be linked to detailed informa-

tion about the investor. Attributes reported in FCSD include the investors

type ~household or institution, with the latter further broken down into

four types of institutions!, native language,9 and municipality. We measure

the distance between a reporting investor and the firm as the distance in

meters between the centroid of the investors municipality and the centroid

of the municipality of the firms headquarters ~generously computed for us

by respected Finnish researchers in geography!.10

6 Annual reports provided geographic location, CEO name, and firm language when Prssi-

tieto did not contain them.7 A previous draft of this paper, with similar language results, classified multilingual firms

that did not use Swedish in their annual report as Finnish-only. Because the current draft now

includes a specific distinction between language and culture, and because many Finns, irrespec-

tive of mother tongue, speak other languages, particularly English, we reclassified all multi-

lingual firms together to isolate the impact of communication per se on investor preferences for

holding and trading stock.8

If either the first or last name of the CEO is of Finnish origin, we classify the CEO ~andfirm! as being of Finnish cultural origin. If the CEOs first and last names are both of Swedish

origin, we further investigate the native tongue of the CEO for the cultural classification. For

this, we checked four Finnish biographical sources ~Whos Who in Finland 1998, and three

listings of Finnish university graduates! for a reported mother tongue of the CEO. In a few

instances, we inferred the mother tongue from the language of his university education. The

results are essentially unchanged if the board chairmans cultural origin is used instead of the

CEOs cultural origin.9 We exclude the fewer than 0.1 percent of Finnish shareowners whose mother tongue is

neither Swedish nor Finnish. This excludes only 546 of almost 500,000 buy transactions and

proportionately fewer sells and shareownership data points.10 If the investor lives in the same municipality as the firm, we define distance as one

quarter of the distance between the centroid of the municipality and the nearest neighboring

municipality. As Thomas and Huggett ~1980, p. 137! note, this convention is customary in the

literature that models geographic phenomena.



5/21

B. Data Aggregation

Because of the exceptional sample size of the data on the holdings and

transactions of investors, it is computationally infeasible to perform statis-

tical analysis on the cross section at the investor level. For this reason, weaggregate our data on investors at the municipality level, which leaves us

with 44,135 firmmunicipality combinations. This aggregation is done sep-

arately for households and institutions, as well as for the institutional sub-

sets of financially savvy institutions ~non-financial corporations, finance and

insurance institutions!, which constitute the vast majority of the institu-

tions, and unsavvy institutions ~government, nonprofits!. Later in the paper,

we also separately aggregate data for household investors who have similar

numbers of distinct stocks in their portfolio. Thus, for each municipality in

Finland, we compute across all investors of a given type:

1. the number of shareowners in the municipality in each of the 97 Fin-

nish firms in the FCSD;

2. the aggregate number of buy ~sell! transactions ~irrespective of the

number of shares! in the municipality for each FCSD Finnish firm

over the approximate two-year sample period; and

3. the fraction of shareowners ~buyers, sellers! in the municipality whose

mother tongue is Swedish, both for the aggregated shareowners ~buyers,

sellers! in the municipality and for the shareowners ~buyers, sellers! of

each firm. Note that because of the summation, investors who own

shares in n firms have their ownership counted n times in both thenumerator and denominator of this fraction.

II. Holdings, Purchases, and Sales as a Function

of Distance, Language, and Culture

A. Some Simple Ratios That Document Distance Effects

Table I, Panel A reports ratios that document the influence of distance on

the behavior of Finnish investors. Two columns, representing households

and institutions, generate two ratios each for shareowners ~first two col-umns!, buys ~middle two columns!, and sells ~final two columns!.

The first two columns in Panel A report statistics on the ratio

Firm i s shareowner weight for investorsin the municipality of its headquarters

Firm i s shareowner weight among all investors in Finland.

The numerator is simply the number of household shareowners of firm i

residing in the municipality the firm is headquartered in, divided by thesum, across all firms, of the number of shareowners residing in that same



6/21

Table I

Summary Statistics Documenting the Influence of Distance, LanPanel A reports the median and fraction exceeding one on the ratio of firm i s weight for investors in tfirm i s weight among all investors in Finland. Panels B and C report the median and fraction exceed

weight among Swedish-speaking investors over firm i s weight among all investors. Panel B focuses o

language of the firm, and Panel Cs subsamples are generated by the cultural origin of the CEO. We

owners, buys, and sells, and are separately computed for households, institutions, and for Panel A, bo

firms and non-Helsinki firms. The number of shareowners is computed as of January 1, 1997, whereas

December 27, 1994, and January 10, 1997.

Weights Based on

Shareowners Buys

HouseholdsAll

Institutions Households I

Panel A: Distance Effect

Summary Statistics for the Ratio Numerator/Denominator

Numerator Firm i s Weight among Investors in its Municipali

Denominator Firm i s Weight among All Investors in Finland

Median for firms of following type

Helsinki area headquartered companies ~n 62! 1.41 1.26 1.11

Rest of Finland headquartered companies ~n 35! 12.16 8.12 4.22

All companies ~n 97! 1.81 1.45 1.34

Fraction greater than 1 for firms of following type

Helsinki area headquartered companies 0.83 0.78 0.71 Standard error 0.07 0.07 0.06

Rest of Finland headquartered companies 1.00 0.91 0.89

Standard error 0.09 0.09 0.08

All companies 0.89 0.83 0.77



7/21

Panel B: Language Effect

Summary Statistics for the Ratio Numerator/Denominator

Numerator Firm i s Weight among Investors who Speak Swed


Median for firms of following typeAnnual report only in Swedish n 2 7.77 4.12 4.84

Annual report only in Finnish n 12 0.52 0.43 0.59

Annual report multilingual n 83 0.85 0.98 0.92


Annual report only in Swedish 1.00 1.00 1.00


Annual report only in Finnish 0.00 0.10 0.17


Annual report multilingual 0.35 0.46 0.43


Panel C: Culture EffectSummary Statistics for the Ratio Numerator/Denominator

Numerator Firm i s Weight among Investors who Speak Swed


Median for firms of following type

CEO Swedish culture n 14 2.49 1.82 1.73

CEO Finnish culture n 83 0.77 0.87 0.86


CEO Swedish culture 0.71 0.79 0.79


CEO Finnish culture 0.27 0.38 0.35



8/21

municipality. The denominator is the comparable ratio for all of Finland. As

an example, take the shipping firm Birka Line, which has 3,299 household

shareowners, 1,669 of whom live in its headquarters city of Mariehamn.

Summing the number of household shareowners over all firms, we find thatMariehamn has 14,440 household shareowners, while Finland has 1,157,783

shareowners. The numerator for Birka Lines ratio is thus 1,669014,440, and

the denominator is 3,29901,157,783, making Birka Lines ratio 40.56.

The two types of summary statistics for this ratio are the median across

firms and the fraction greater than one. In the absence of a distance effect,

this ratio is one.11 However, the third row of Panel A indicates that the me-

dian ratio for households is 1.81 and for institutions it is 1.45. Restricting

the set of firms to those headquartered outside of greater Helsinki makes

the distance effect stand out even more. The median ratio for households is

12.16 and for institutions it is 8.12. For the more nationally known compa-nies headquartered in the greater Helsinki area, the median ratios are a

more modest 1.41 and 1.26 for households and institutions, respectively.

The fraction of firms with ratios that exceed one is equally impressive: The

first column of Table I, Panel A, indicates that 100 percent of the 35 non-

Helsinki firms and 83 percent of the Helsinki-area firms have household

shareownership ratios that exceed one. Again, institutions, seen in the sec-

ond column, indicate a more modest distance effect, but a strong one none-

theless. Using the standard errors in the table, it is evident that both

percentages are statistically significant. Also, although unreported, the dif-

ferences between the Helsinki and non-Helsinki firms in the fraction of the

ratios that exceed one are statistically significant, except for institutional

shareownership and buys.

The results in the four rightmost columns show that distance strongly

influences the active purchases and sales of seasoned stock for non-Helsinki

firms. The median ratios here, using the number of buy transactions, but

computed analogously to the ratios for shareownership, are smaller than

those for shareowners in the first two columns, but are well above one and

exhibit the same pattern across the subcategories of investors. The same is

true of the fraction of buy ratios that exceed one. In contrast to the share-

ownership results, institutions do not seem to exhibit distance effects in

their trading of Helsinki-area companies.

Table I, Panel A, has clearly documented that distance influences investor

behavior. We will later show that this distance effect does not arise from the

ownership, purchases, and sales of customers and0or employees of the firm,

who would naturally tend to reside close to it, nor is it due to language or

culture effects that are linked to where investors live and firms locate.

11 The expected numerator is the same as the expected denominator under the null. How-

ever, because of Jensens inequality, the expectation of the ratio is greater than one. For this

reason, as well as the effect of outliers, we report only the median and fraction exceeding one,

although the results with the mean are very similar.



9/21

B. Some Simple Ratios That Document the Influenceof Language on Investor Behavior

The first two columns in Table I, Panel B, report summary statistics on

the ratio

Firm i s shareowner weight among Swedish-speaking investors

Firm i s shareowner weight among all investors in Finland.

The denominator is identical to that used in the prior subsection to analyze

distance. The numerator is simply firm i s number of Swedish-speaking share-

owners divided by the sum over all firms of the number of shareowners. The

ratio is separately computed for household and institutional investors.

To illustrate this ratio, let us return to the example of Birka Line, a pure

Swedish-speaking company, which ~you may recall! has 3,299 household share-owners, 3,147 of whom report that Swedish is their native language. Be-

cause there are 127,750 Swedish-speaking household shareowners in all of

Finland and 1,157,783 household shareowners overall, the numerator for

the ratio is 3,1470127,750 and the denominator is 3,29901,157,783, making

Birka Lines ratio 8.65.

In the absence of a language effect, this ratio is one. However, the first

row of Panel B indicates that the median ratio is 7.77 for household share-

owners and 4.12 for institutional shareowners among the Swedish-only firms,

and the second row of Panel B indicates that the same ratios are, respec-

tively, 0.52 and 0.43 among the Finnish-only companies. In other words, themedian Swedish language firm is almost 15 times more popular than the

median Finnish language firm among Swedish-speaking household inves-

tors. Even for institutions, this difference is almost tenfold. The reverse is

necessarily true for Finnish-speaking investors.

The fraction of firms with ratios exceeding one, as well as the median and

fraction greater than one for analogous ratios computed with buys and sells,

tell a similar story. Finnish-speaking investors prefer Finnish-speaking firms,

Swedish-speaking investors prefer Swedish-speaking f irms, and multilin-

gual firms lie somewhere in between these two extremes in their relative

proportions of Swedish-speaking and Finnish-speaking investors.12 Note thatbecause the ratios are based on differences between Finnish-speaking inves-

12 The results are supported by evidence on foreign investment in Finland, which we do not

formally report for brevitys sake. In particular, there is a large community of former Finnish

residents in Sweden, consisting mostly ~but not entirely! of Finnish-speaking people who have

moved to Sweden during the last few decades. Comparing the Finnish investments of these

Finnish-speaking Swedes ~largely, the expatriate Finns! with the Finnish investments of Swedish-

speaking Swedes tells a similar story to that in Panel B. For example, there are two Finnish

firms that report only in SwedishBirka Line and Chips. The former has 557 native Swedish

speakers among its 562 Swedish investors ~99 percent!. For Chips, the respective ratios of

native Swedish speakers are 88 out of 93 among its Swedish investors ~95 percent!. By contrast,

the aggregate Finnish stock market has only 74 percent Swedish speakers among its Sweden-

domiciled investors.



10/21

tors and Swedish-speaking investors, it is impossible to determine which of

the following is the ultimate source of the observed language effect: ~1! Finnish-

speaking investors preference for the pure Finnish language companies;

~2! Swedish-speaking investors preference for the pure Swedish languagecompanies; ~3! Finnish-speaking investors shunning firms with a pure Swed-

ish language orientation; or ~4! Swedish-speaking investors shunning the

pure Finnish language firms. As with distance effects, institutional inves-

tors appear to be less influenced by the language of the firm than household

investors.

C. Some Simple Ratios That Document the Effectof Culture on Investor Behavior

The first two columns in Table I, Panel C, report summary statistics on

the same ratio reported in Panel B. Here, however, we subdivide firms basedon firm culture. The first row of Panel C indicates that the median Swedish-

speaking shareownership ratio among the Swedish-culture firms is 2.49 for

households and 1.82 for institutions. The second row of Panel C indicates

that the same ratios are, respectively, 0.77 and 0.87 among the Finnish-

culture companies. The fraction of firms with ratios exceeding one, as well

as the median and fraction greater than one for analogous ratios computed

with buys and sells, tell a similar story of preference for firms of the same

culture.13 As with language and distance, institutions are less inf luenced by

the culture of the firm than households.

D. Multivariate Regression Motivation,Variable Description, and Results

One difficulty in interpreting Table Is results is that investors with sim-

ilar language and culture tend to live near one another. Hence, as noted in

the introduction, a preference for proximate investments may be a manifes-

tation of language or culture effects or vice versa: Finnish firms that tend to

communicate in Swedish or have a Swedish cultural origin may locate near

Swedish investors.

To disentangle distance effects from language or cultural effects, and tocontrol for other potentially confounding variables, this subsection analyzes

the determinants of a dependent variable Dij , the difference between mu-nicipality j s weight on firm i, Xij , and the markets weight on firm i, Xi .That is,

Dij Xij Xi .

13 The culture effect also applies to investors in Finnish stocks who are domiciled in Sweden.

Controlling for the firms language in an unreported dummy variable regression, Swedish-

speaking investors domiciled in Sweden are 22.4 percent more likely to hold shares in a Finnish

company whose CEO is of Swedish cultural origin ~t 2.44! than a company headed by a CEO

of Finnish cultural origin.



11/21

Thus, Dij is the difference between the numerator and the denominator usedto compute the ratios in Table I, Panel A, except municipality j can be any mu-

nicipality rather than just the municipality in which firm i is headquartered.

Table II is based on regressions in which the dependent variable, Dij , isprojected onto ~1! dummy variables for each stock ~but one!; ~2! dummy vari-

ables for each municipality ~but one!; ~3! the maximum of 100 kilometers

and the log of the distance ~in meters! of municipality j from the headquar-ters municipality of firm i;14 ~4! the minimum of 100 kilometers and the logof the distance of municipality j from the headquarters municipality of firm

i; ~5! two distance slope dummies, to ascertain whether a firm that is head-quartered in the greater Helsinki area has a different ~distance! coefficient

for items 3 and 4 above than other firms; ~6! two language variables repre-

senting the products of dummy variables associated with the language of the

annual report times the fraction of Swedish-speaking shareowners ~buys,sells! in the municipality; and ~7! a Swedish culture variable that is the

product of a dummy variable that is one if the CEO is of Swedish cultural

origin times the fraction of Swedish-speaking shareowners ~buys, sells! in

the municipality. The regressions are separately run for households and in-

stitutions, as well as for two institutional subcategories, which group pairs

of institutional types into financially savvy and unsavvy institutions.

The piecewise linear functional form of the distance regressors, which em-

ploy a single change in slope at 100 kilometers, is inspired by Figure 1. It

plots coefficients representing the marginal effect on shareownership, Dij , ofa group of distance interval dummy variables from 0 to 450 kilometers, with

each interval representing 5 kilometers. The regression used to obtain the

plotted dummy coefficients also controls for the variables described above,

except for distance. Panel A in Figure 1 plots household shareowner coeffi-

cients for the distance dummies, using both Helsinki firms ~distance dum-

mies plus Helsinki slope dummies! and non-Helsinki f irms ~distance dummies

alone!. Panel B plots comparable institutional shareowner coefficients.

Table II reports the coefficients and t-statistics for a constant and 7 regres-sors in each of 12 regressions with Dij as the dependent variable. The tabledoes not report the regression coefficients on about 550 other included regres-

sorsdummy variables for the firms and municipalitiesthat control for cross-

sectional persistence in shareownership, buys, and sells due to omitted variables

like firm size or urbanrural differences in familiarity with the stock market.

All 12 of the regressions indicate that there is a strong distance effect for

non-Helsinki firms with a larger marginal distance effect below 100 kilometers.

For the under-100 kilometer distances, the t-statistics range from 6.00 to40.92. The marginal effect of distance for non-Helsinki firms beyond 100

kilometers is negligible for the institutional buy transactions, but is other-

wise significant.

14 The log functional form for distance is inspired by other distance literature in the social

and physical sciences @see, for example, Sheppard ~1984!#. However, the Table II results are

virtually the same whether the regressors are based on the log of distance or unlogged distance.



12/21


13/21

Figure 1. Graphs of coefficients on distance dummy variables in a regression that is

identical to that in Table II for shareownership, except that the two distance re-

gressors are replaced by two sets of dummy variables representing five-kilometer

distance intervals from 0 to 450 kilometers. Separate sets of dummies for Greater Hel-

sinki Area headquartered firms and other firms are plotted. The dependent variable in the

regression is firm i s weight for investors in a given municipality less firm i s weight among

all investors in Finland. Each data point is a firm-municipality combination. Panel A plots

the coefficients on the dummy variables for households and Panel B plots the coefficients

on the dummy variables for institutional investors. Shareownership is computed as of Janu-

ary 1, 1997.



14/21

The positive coefficients on the Helsinki slope dummies also indicate that

the more prominent Helsinki-headquartered firms have negligible if any

distance effect exhibited towards them.15 The reduction in the distance effect

exhibited towards Helsinki-headquartered firms is not a size effect. An un-reported regression that adds two more distance cross-product variables,

computed analogously to the Helsinki variables but with the log of the num-

ber of employees replacing the Helsinki dummies, has very similar results;

the number of employees does not moderate the distance effect.

The lack of an effect of the number of employees on the distance effect also

suggests that the distance effect results in Table II are not due to employees

owning, buying, or selling their own firms stock. Only one of the 24 employee-

related distance variables in the 12 regressions is significant at the five

percent level. Buttressing this argument is the functional form of the dis-

tance effect, which, if due to employee ownership, should be bunched nearzero rather than be piecewise linear, as well as the similar results obtained

from regressions that exclude all investors from the firms home municipal-

ity ~not formally reported for the sake of brevity!. Finally, the fact that not

just households, but institutions, too, are inf luenced by distance suggests

that this effect may be part of a behavioral phenomenon that is not tied to

the behavior of the firms employees.

Consistent with the results in Table I, distance influences the investment

behavior of institutions much less than households. For example, the house-

hold shareownership regression has an under 100 kilometer distance coef-

ficient of0.024, whereas the institution shareownership regression has a

comparable coefficient of0.016. The coefficient on the above 100 kilometer

distance regressor is twice as large for households as for institutions. This

pattern, in which households exhibit larger distance effect, exists for buys

and sells as well.

If greater investor sophistication accounts for the relatively smaller effect

of distance and language on the investment behavior of institutions, then

distance may have less of an impact of the behavior of the most financially

savvy institutions. Table II suggests that this is indeed the case. For the

firms headquartered outside of greater Helsinki, five of the six institutional

distance coefficients are significantly negative. In each of these five signif-

icant cases, a comparison of coefficients indicates that there is less of a

distance effect for the savvy institutionsnon-financial corporations and fi-

nance and insurance institutionsthan for the unsavvy institutions. Indeed,

the smaller influence of distance on all institutions relative to households

appears to be entirely driven by the financially savvy institutions. Because

most of the unsavvy institution category consists of nonprofits, we suspect

that what drives the larger distance effect for the unsavvy institutions is the

tendency of charitable foundations, particularly those located far from Hel-

sinki, to invest heavily in local firms.

15 The lone exception to this, government and nonprofit buys when such institutions are

located far from Helsinki, is statistically insignificant.



15/21

The regression also indicates that in addition to distance, both language

and culture influence the investment behavior of households. Swedish-

speaking household investors have a stronger desire to own and buy com-

panies that communicate exclusively in Swedish. Finnish-speaking investorshave a relatively greater aversion to these Swedish-language companies. How-

ever, the language of the household investor does not generate a relative pref-

erence among Swedish-speaking investors for multilingual companies over

Finnish-only companies. With one exception on ownership, there does not ap-

pear to be much evidence of a language effect among institutions. We believe

that these nonresults are due to the low power of this regression specifica-

tion to pick up language effects, as we will demonstrate at the end of this section.

Finally, there is quite a strong relation between culture and investor pref-

erences, particularly for households. Swedish-speaking households prefer

~whereas Finnish-speaking households disprefer! to hold and trade the sharesof companies with CEOs of Swedish cultural origin, controlling for language

and distance. The shareownership of institutions also is modestly altered by

the culture of the firm.

E. More Evidence on Sophistication and Distance Effects

We have so far demonstrated that distance inf luences the behavior of

investment-savvy institutions less than the behavior of households and less

savvy institutions, suggesting that there may be a link between the sophis-

tication of the investor type and the influence of distance. To investigate this

link further, we analyze the impact of the number of stocks held on the

distance coefficients. We first separate our sample of households into 10

different subgroups based on number of different stocks held1 through 9,

and 10 or greaterand then aggregate these subgroupings of household in-

vestors into firmmunicipality combinations. Table III repor ts the results of

the same regressions in Table II on these subgroups of household investors.

Although all of the subsamples exhibit distance effect, the inf luence of dis-

tance, for the most part, appears to be smaller the greater the number of

different stocks held. To some extent, this relation between sophistication

~as proxied for by different number of stocks held! and distance effects is a

manufactured result, in that if investors have lexicographic preferences basedon distance from a firm, those with more diversified ~and, consequently,

more sophisticated! portfolios necessarily hold more distant firms. However,

this interpretation of Table III still implies that there is a link, albeit an

indirect one, between distance effects and sophistication.16 There is also a

16 This link is also supported by three ~unreported! regressions for households that add distance

income ~based on the municipalitys average income per household! and distanceeducation ~based

on the fraction of adults in the municipality with a high school diploma or equivalent! interaction

terms to the regressors in Table II. The four interaction terms are generally significant in all three

regressions. For proximate firms ~closer than 100 kilometers!, the income interaction term t-statistics

range from a marginally significant 1.95 to a highly significant 5.86. Also, for proximate firms,

the education interaction term has t-statistics that range from 4.56 ~buys! to 17.19 ~shareowners!.



16/21

Table III

Multivariate Regressions that Analyze the Relation between Port

and the Influence of Distance, Language, and Cul

Table III reports goodness of fit, number of observations, and coefficients ~with t-statistics below! for a 10 regressions with the dependent variable being firm i s shareowner weight for investors in a given

distinct stocks in their portfolio less firm i s shareowner weight among all investors in Finland with the

is a firmmunicipality combination. The table does not report the regression coefficients on about

variables for the firms and municipalities. The number of shareowners is computed as of January 1,

Dependent Variable: Weight on Firm i in Municipality j Less Overall Weight on Firm i with Weight

Number of Stocks in Po

Independent Variables 1 2 3 4 5 6

Constant 0.068 0.113 0.097 0.091 0.074 0.07

6.06 9.64 8.95 8.31 6.76 6.73

Min @ln 100, ln ~distance!# 0.022 0.034 0.027 0.026 0.022 0.01

26.88 39.41 33.84 31.96 27.08 21.08

Max @ln 100, ln ~distance!# 0.004 0.008 0.008 0.007 0.006 0.00

7.53 13.38 13.46 12.79 10.39 9.67

Min @ln 100, ln ~distance!# Dummy for company 0.020 0.030 0.023 0.022 0.019 0.01

headquartered in Greater Helsinki Area 16.36 23.31 19.72 18.67 15.82 12.78

Max @ln 100, ln ~distance!# Dummy for company 0.001 0.002 0.003 0.003 0.002 0.00

headquartered in Greater Helsinki Area 1.09 3.64 4.47 4.39 3.85 1.53

Fraction of Swedish speakers in municipality 0.002 0.000 0.001 0.001 0.001 0.00

Annual report only in Finnish dummy 1.52 0.27 0.41 0.35 0.41 0.54

Fraction of Swedish speakers in municipality 0.055 0.034 0.007 0.023 0.033 0.02

Annual report only in Swedish dummy 14.19 8.54 1.80 6.37 8.96 6.33

Fraction of Swedish speakers in municipality 0.012 0.033 0.019 0.019 0.016 0.01CEO Swedish culture dummy 7.61 21.04 13.07 13.25 10.63 10.15

Adjusted R2 0.275 0.137 0.113 0.087 0.064 0.04

N 41676 41676 41492 40296 38916 3440


17/21

similar link between cultural effects and the number of stocks in the house-

hold investors portfolio, which can be observed in Table III.

F. Motivation for Further Analysis of Languageand Culture Effects and Regression Results

Table IIs analysis of language effects examines how the fraction of inves-

tors who speak Swedish in a municipality affects the municipalitys overall

propensity to hold, buy, or sell a company that communicates exclusively in

Swedish, exclusively in Finnish, or in multiple languages. It should not be

surprising that such tests fail to pick up language effects, as was generally

true when comparing multilingual firms with Finnish-only firms. The de-

pendent variable in Table II captures above- ~or below-! normal investmentactivity in a given firmmunicipality combination. Such abnormal invest-

ment activity is largely generated by omitted factors ~although we have firm

and municipality dummies to control for them! and noise. While municipality-

to-municipality variation in the fraction of Swedish speakers may be a con-

tributor to abnormal investment activity in a firm, it cannot account for

much of the variation in abnormal investment activity when Finnish-

speaking investors dominate ownership and trading in all but a few of Fin-

lands 455 municipalities.

As an example of this problem, and how to remedy it by altering the depen-

dent variable, take the municipality of Mariehamn. This Swedish-speaking

community may be relatively more invested in the two Swedish-language-

only companies than in the multilingual companies and more invested in

the multilingual companies than in Finnish-language-only companies.

Similarly, it may be more invested in the Swedish-culture companies. How-

ever, pooling the Swedish-speaking investors with Finnish-speaking inves-

tors in Mariehamn, as we do in Table II, does not fully take advantage of

the knowledge of the allocation of Swedish-speakers investments into dif-

ferent types of companies. Regressing the fraction of Swedish-speaking own-

ers of a company in a given municipality, like Mariehamn, on dummies

representing the language~s! of the companys annual report and the cul-

tural origin of the CEO would pick these effects up better, even within

Mariehamn. It also can better capture any culture effects, although there

has been ample evidence of this, even within the lower power specification

of Table II.

Table IV thus reports coefficients and t-statistics for regressions of thefraction of Swedish-speaking shareowners in a given firmmunicipality

combination against dummies for whether the annual report of the company

is exclusively in Finnish or exclusively in Swedish ~the default dummy being

a multilingual annual report!, as well as the Swedish culture dummy gen-

erated by the cultural background of the firms CEO. As controls, the re-

gressions also include the four distance variables used in Table II, a dummyfor companies headquartered in the Greater Helsinki Area, as well as



18/21

unreported intercept dummies for each municipality. ~Obviously, company

dummies cannot be used here as they subsume the dummies which control

for the language of the annual report.!17

17 While the distance variables control for the distance effect of similar-language investors

living near similar-language companies, which would be accidentally picked up as a language

effect if the distance controls were omitted, their coefficients are not good estimates of distance

effect per se.

Table IV

Multivariate Regressions that Analyze the Determinants

of the Language and Culture of Investors

in a FirmMunicipality CombinationTable IV reports goodness of fit, number of observations, and coefficients ~with t-statistics

below! for a constant and eight regressors. The dependent variable is the fraction of Swedish-

speaking shareowners. Unreported dummies for each municipality are also included. The re-

gressions are separately computed for households, institutions, savvy institutions, and unsavvy

institutions. Each data point is a firmmunicipality combination. The number of shareowners

is computed as of January 1, 1997.

Dependent Variable:The Fraction of Firm i s Shareowners Residing

in Municipality j who Speak Swedish

Institutions

Independent Variables Households All

Non-Fin.Corp & Fin.

and Ins. Inst.

Government& NonprofitInstitutions

Constant 0.067 0.097 0.075 0.1051.55 1.26 0.92 0.60

Min @ln 100, ln ~distance!# 0.005 0.008 0.006 0.0181.31 1.31 0.90 1.72

Max @ln 100, ln ~distance!# 0.009 0.016 0.012 0.0252.98 2.38 1.68 1.48

Min @ln 100, ln ~distance!# Dummyfor company headquartered in

Greater Helsinki Area

0.005 0.005 0.002 0.0190.94 0.56 0.22 1.37

Max @ln 100, ln ~distance!# Dummyfor company headquartered inGreater Helsinki Area

0.011 0.010 0.012 0.0053.04 1.24 1.33 0.23

Dummy for company headquarteredin Greater Helsinki Area

0.193 0.068 0.122 0.2783.32 0.68 1.14 1.22

Annual report only in Finnish dummy 0.020 0.011 0.005 0.0656.43 1.70 0.73 3.13

Annual report only in Swedish dummy 0.062 0.009 0.020 0.0638.29 0.52 1.07 1.34

CEO Swedish culture dummy 0.025 0.032 0.025 0.03511.32 6.76 4.87 3.83

Adjusted R2 0.834 0.824 0.817 0.770

N 21279 7612 6928 2521



19/21

Table IV indicates that there is a much stronger language inf luence and similar

~or slightly more! cultural influence than was evident from the shareowner-

ship regressions analyzed in Table II. For households, the fraction of Swedish-

speaking shareowners in Swedish-language firms is a statistically significant0.062 higher than that in a multilingual company and 0.082 higher than in a

pure Finnish-language f irm, ceteris paribus. Culture adds to this effect: The

fraction of Swedish-speaking household shareowners in a firm of Swedish cul-

tural origin is 0.025 larger than a firm with CEO of Finnish cultural origin.

For institutions, the respective increase in fractional Swedish-speaker

ownership for Swedish-speaking firms is a modest 0.009 relative to multi-

lingual firms and 0.020 relative to Finnish-only firms. For financially un-

savvy institutions, however, the respective increases in fractional Swedish

ownership are 0.063 and 0.128, respectively, which are much larger. This

indicates that the more modest influence of language on the investment be-havior of institutions is driven by the savvy institutions, as was the case with

distance. Because most of the unsavvy institution category consists of non-

profits, we suspect that what drives the larger influence of language on the

behavior of unsavvy institutions is the tendency of Swedish charitable foun-

dations to invest heavily in Swedish language and multilingual companies.18

III. Summary and Conclusion

This paper documents that investors simultaneously exhibit a preference

for nearby firms and for same-language and same-culture firms. We presenta substantial amount of evidence that seems to support the hypothesis that

the degree of these effects is inversely related to investor sophistication.

The language results are particularly unique and they may have implica-

tions for firm policy. Finland-domiciled companies that publish their annual

reports both in Finnish and Swedish are able to tap an abnormally large

Swedish-speaking investor base, both in Finland and Sweden. Firms in other

countries should be able to do the same to increase their investment appeal.

For example, U.S. companies, which generally publish their annual reports

only in English, might be able to expand their investor base by publishing

their annual reports also in, say, Spanish and Japanese.

In Europe, the success of the merging of the national stock exchanges ~see

Andrews ~1999!! will be affected by how much distance, language, and cul-

tural preferences alter the f low of investment capital between countries. The

latter two barriers to inter-European investment are more difficult to over-

come. However, according to the results in this paper, this experiment is

more likely to succeed in altering the home bias already exhibited in Europe

if companies listing in the unified stock market can overcome language bar-

riers in their communications to investors, particularly for investors in nearby

countries and in countries that share the same culture as the firm.

18 The influence of language and culture on the shareownership of unsavvy institutions does

not carry over to the trades of these institutions, which is consistent with the findings of Table II.



20/21

We believe that our results on the influence of language, distance, and

culture are fairly robust. Although our analysis, whether focused on buys,

sells, or shareownership, gives the same weight to each investor ~or trans-

action! irrespective of the size of the investment, we have checked that theresults are similar when we weight investors by the size of their invest-

ment. However, we cannot be certain that the behavioral results here will

carry over to larger financial markets like the United States or United King-

dom. This distinction is important because the vast majority of Finnish in-

vestors hold poorly diversified portfolios. Ilmanen and Keloharju ~1999!, for

example, report that a household investor holds an average of two stocks.

This is not very different from shareholders in the United States about 20

years earlier, but it excludes what appears to be the U.S. investors more

diversified holdings of stocks by virtue of mutual fund and pension fund

investment.Despite the precedent in the home bias literature, we have been careful to

this point not to classify the influence of distance, language, and culture as

biases, which connotes that some form of investor irrationality is behind

the influence of these factors. For portfolios that are as poorly diversified as

those of most households in Finland, the biases that have been identified

here have little effect on the risk profile of the investors holdings. The dam-

age has been done by poor diversification per se. However, for an investor

who chooses to hold a large number of stocks, concentrating the portfolio in

certain stocks because of distance, language, or culture effects may make

quite a large difference to the risk profile of his investment holdings. Con-

sistent with this cost, we find more modest evidence of such effects among

institutions and those households with larger numbers of firms among their

holdings. However, the existence of such effects at all among the more so-

phisticated Finnish investors, as well as their lesser influence among the

more sophisticated investor groups, leads us to conjecture that investors

generally prefer to hold and trade stock in more familiar firms. If our con-

jecture is correct, then the investment regularities exhibited towards famil-

iar firms in Finland probably exist in other countries, even among those

with more diversified holdings. It would naturally follow that such familiarity-

related effects could be the major contributor to home bias.

Of course, it is possible that any familiarity bias could be rational. In-

vestors may acquire useful information about familiar firms from reading

company statements in a language they understand, f rom general or ac-

quired knowledge about local firms, or from the cultural groups they social-

ize within. Such an information-based theory of the influence of distance,

language, and culture would be manifested in more active trading of these

familiar firms and would generate superior performance in these firms. A

performance test to verify this is beyond the scope of this paper. However,

because Grinblatt and Keloharju ~2000! have confirmed that portfolio per-

formance in Finland is inversely related to investor sophistication, we are

skeptical about superior information as the source of the influence of famil-iarity on both holdings and trades.



21/21

It is also possible but unlikely that our results are driven by reverse cau-

sation. For example, firms may choose to report in the language of their in-

vestors or their potential investors. Also, the f irms investors choose the board

of directors who in turn choose the CEO. However, we think this alternativeexplanation is unlikely for two reasons. The first reason is the strength and

relative strength of the buy results, which are the least influenced by prior ac-

tual or forecastable relationships with the firm. Except for distance, the buy

results in Tables I and II are about as strong as the sell results and never less

than one-half the strength of the shareownership results ~where prior actual

relationships are known to exist!. It is also easy to be skeptical about the hy-

pothesis that f irms would choose to locate their headquarters in more remote

areas of Finland to be in close proximity to potential buyers of the firms stock.

Second, the reverse causation hypothesis would imply that results that weight

investors by the market capitalization of their portfolios should be stronger thanthose presented here. Instead, they are similar or somewhat weaker than those

that identically treat each investor or each trade.

REFERENCES

Andrews, Edmund, 1999, 8 European stock exchanges form alliance, New York Times, May 4.

Cooper, Ian, and Evi Kaplanis, 1994, Home bias in equity portfolios, inflation hedging and

international capital market equilibrium, Review of Financial Studies 7, 4560.

Coval, Joshua, and Tobias Moskowitz, 1999a, The geography of investment: Informed trading

and asset prices, Working paper, University of Michigan.

Coval, Joshua, and Tobias Moskowitz, 1999b, Home bias at home: Local equity preference in

domestic portfolios, Journal of Finance 54, 20452073.French, Kenneth, and James Poterba, 1991, Investor diversification and international equity

markets, American Economic Review 81, 222226.

Grinblatt, Mark, and Matti Keloharju, 2000, The investment behavior and performance of various

investor-types:A study of Finlands unique data set,Journal of Financial Economics 55, 4367.

Grinblatt, Mark, and Matti Keloharju, 2001, What makes investors trade?, Journal of Finance

56, 589616.

Huberman, Gur, 1998, Familiarity breeds investment, Working paper, Columbia University.

Ilmanen, Matti, and Matti Keloharju, 1999, Shareownership in Finland, Finnish Journal of

Business Economics 48, 257285.

Kock, Gunhard, 1995 and 1996, Prssitieto ~Prssitieto Ky, Helsinki, Finland!.

Kang, Jun-Koo, and Ren Stulz, 1997, Why is there a home bias? An analysis of foreign port-

folio equity ownership in Japan, Journal of Financial Economics 46, 328.Petersen, Mitchell, and Raghuram Rajan, 2000, Does distance still matter? The information

revolution in small business lending, Working paper, Northwestern University.

Serrat, Angel, 1997, A dynamic equilibrium model of international risk-sharing puzzles, Work-

ing paper, University of Chicago.

Sheppard, Eric, 1984, The distance-decay gravity model debate; in Gary Gaile and Cort Will-

mott, eds.: Spatial Statistics and Models ~D. Reidel Publishing Company, The Netherlands!.

Stulz, Ren, 1981a, On the effect of barriers to international investment, Journal of Finance 36,

923934.

Stulz, Ren, 1981b, A model of international asset pricing, Journal of Financial Economics 9,

383406.

Tesar, Linda, and Ingrid Werner, 1995, Home bias and high turnover, Journal of International

Money and Finance 14, 467492.

Thomas, Reginald W., and R. J. Huggett, 1980, Modelling in Geography. A Mathematical Ap-

proach ~Barnes & Noble Books, Totowa, New Jersey!.


grinblattk_jf56

Documents